Performance improvement of the license plate recognition systems with algorithmic fusion approach

Abstract

In this paper, we propose an algorithmic fusion approach for performance improvement of the license plate recognition (LPR) systems which is a very important research topic in computer vision and pattern recognition areas. Basically, the building blocks of LPR system is composed of locating the LP, character segmentation and character recognition stages. Character segmentation and recognition phases are well studied in the fields of image processing and pattern recognition and very robust solutions were proposed in the literature. However, the most important problem of the LPR systems is locating the pixel precise position of the LP inside the image. In our previous work, we proposed a horizontal scan method to locate the LP. In this paper, we propose a square detection based LP locating algorithm which also locates the two lined LPs. In order to design a robust stage for the locating plate block, we present a new approach that combines the dexterity of the two different LPR locating algorithms within algorithmic fusion framework. According to our experiments, we observe that the overall performance of the LPR system is increased by the proposed algorithmic fusion method.